Nvidia分潤協議評測:360億美元AI雲端合作喊停 | Nvidia Review: $36B AI Cloud Revenue-Share Deal Paused
By Kit 小克 | AI Tool Observer | 2026-08-31
🇹🇼 Nvidia分潤協議評測:360億美元AI雲端合作喊停
Nvidia分潤協議上週被爆出緊急喊停——這家全球市值最高的晶片公司,今年七月才剛端出一套讓AI雲端業者又愛又怕的「分潤協議」,結果不到兩個月,就因為反壟斷疑慮自己踩了煞車。這篇文章帶你搞懂這360億美元的合作葫蘆裡賣什麼藥,以及對正在租GPU算力的中小型雲端商有什麼實際影響。
Nvidia AI雲端分潤協議是什麼?
簡單說,就是Nvidia除了賣晶片賺一次錢,還想在客戶靠這些晶片賺的雲端收入裡再抽一手。根據協議,Nvidia提供「最低營收保證」的信用支持給AI雲端業者,條件是分享超過門檻的營收;如果客戶租不掉算力,Nvidia還承諾自己租回來。等於晶片賣一次、後續營收再抽一次,而且以六年期合約為主。Nvidia這一季財報首度揭露,相關承諾規模高達360億美元。
為什麼突然喊停?反壟斷疑慮出在哪?
問題出在「控制權」。有雲端業者反映,Nvidia要求他們只能把算力租給指定核准的客戶,還偏好把算力平均分給多家中小型AI公司,而不是讓單一大客戶整碗端走。這種對下游客戶生意的介入方式,踩到了反壟斷的紅線——你賣晶片可以,但決定客戶怎麼經營生意就是另一回事。Nvidia內部員工自己先跳出來提醒風險,協議推出不到兩個月就有部分被喊停。
對AI新創和中小型雲端商有什麼影響?
短期來看是鬆一口氣。原本這套分潤協議變相讓Nvidia成為所有AI雲端生意的隱形股東,中小型雲端商簽下去等於把定價權和客戶名單交出去一部分。喊停之後,議價空間回到業者手上,但Nvidia也沒把話說死——未來不排除把這套機制包裝進其他名目重新推出。對正在找GPU算力的團隊來說,值得留意合約裡是否藏著類似的營收綁定條款。
Kit小克怎麼看
老實說,這不是Nvidia第一次在商業模式上玩得比賣晶片更花。從賣鏟子的角色升級成「連挖到的金子都要分一杯羹」,邏輯上很聰明,但風險是把自己從供應商變成整個AI基礎建設的守門人。這次喊停與其說是良心發現,不如說是先避開監管機關的眼睛,這套分潤協議大機率會換個包裝捲土重來。中小型雲端商如果正在跟Nvidia談合約,這幾週會是議價的好時機。
好不好用,試了才知道。
🇺🇸 Nvidia Review: $36B AI Cloud Revenue-Share Deal Paused
Nvidia's AI cloud revenue-sharing deal just hit a wall. Less than two months after rolling out a program that let it collect a cut of AI cloud providers' revenue on top of chip sales, Nvidia has quietly paused parts of it over antitrust concerns. Here's what the $36 billion program actually did, why it got pulled back, and what it means if you're renting GPU capacity from a smaller cloud provider right now.
What Is Nvidia's AI Cloud Revenue-Sharing Deal?
Nvidia offered AI cloud companies minimum-revenue guarantees as credit support in exchange for a cut of revenue above a set threshold. If a customer couldn't rent out its GPU capacity, Nvidia promised to rent it back itself. That means Nvidia got paid twice: once when it sold the chips, and again as a share of whatever cloud revenue those chips generated, mostly under six-year contracts. A quarterly filing this week revealed the total commitments: $36 billion.
Why Did Nvidia Pause the Deal?
Control was the sticking point. Some cloud providers reportedly said Nvidia told them they could only rent capacity to approved customers, and that Nvidia preferred spreading capacity across several smaller AI companies rather than letting one large buyer absorb it all. That level of influence over how customers run their businesses is exactly what draws antitrust scrutiny — selling chips is one thing, dictating downstream business decisions is another. Nvidia's own employees reportedly flagged the risk internally before the pause.
What Does This Mean for AI Startups and Smaller Clouds?
Short term, it's a bit of relief. The revenue-sharing structure effectively made Nvidia a silent stakeholder in every AI cloud deal it touched, with smaller providers giving up pricing power and some customer control in the process. With parts paused, that leverage shifts back to the cloud companies — though Nvidia hasn't ruled out relaunching the program under a different name. If your team is negotiating GPU capacity right now, it's worth checking whether similar revenue-tie-in clauses are buried in the contract.
Kit's Take
This isn't Nvidia's first move to get creative beyond just selling shovels. Going from "sell the pickaxe" to "take a cut of the gold you dig up" is smart business logic, but it also turns a supplier into a gatekeeper for the entire AI infrastructure stack. The pause looks less like a change of heart and more like dodging regulators before they look too closely — expect this deal to resurface in a different wrapper. If you're negotiating with Nvidia right now, this is your window to push back on revenue-sharing terms.
好不好用,試了才知道 — you won't know until you actually try it.
Sources / 資料來源
- Nvidia pauses revenue-sharing deals with AI cloud companies, WSJ reports (Yahoo Finance)
- Nvidia Pauses Revenue-Sharing Deals With AI Cloud Companies Amid Antitrust Concerns (Benzinga)
- NVIDIA Pauses AI Cloud Revenue-Sharing Deals Amid Antitrust Concerns (Open Data Science)
常見問題 FAQ
Nvidia的AI雲端分潤協議是什麼?
Nvidia除了賣晶片賺錢,還向AI雲端業者提供最低營收保證的信用支持,換取分享超過門檻的雲端營收,等於晶片賣一次、後續營收再抽一次。
這套分潤協議規模有多大?
Nvidia最新財報首度揭露,相關承諾規模達360億美元,合約多為六年期,今年七月才剛推出。
Nvidia為什麼喊停分潤協議?
因為要求客戶只能把算力租給核准對象、干預客戶如何分配算力,這種控制下游生意的做法引發反壟斷疑慮,內部員工也提出警告。
對中小型雲端商和AI新創有什麼影響?
短期議價空間回到業者手上,但Nvidia未排除未來把機制包裝成其他名目重推,簽約前建議留意合約中是否有類似的營收綁定條款。
延伸閱讀 / Related Articles
- OpenAI Astra評測:AI資安能力觸頂,訓練緊急喊停 | OpenAI Astra Review: AI Hits Critical Cyber Threshold
- Jalapeño評測:OpenAI自研AI晶片效能超車輝達 | Jalapeño Review: OpenAI's Nvidia-Beating AI Chip
- Perplexity評測:Nvidia砸300億估值投資AI搜尋新星 | Perplexity Review: Nvidia Backs AI Search Startup at $30B
AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends
留言
張貼留言